{"id":3598,"date":"2020-06-25T06:34:35","date_gmt":"2020-06-25T06:34:35","guid":{"rendered":"https:\/\/www.aiproblog.com\/index.php\/2020\/06\/25\/creating-music-by-machine-learning\/"},"modified":"2020-06-25T06:34:35","modified_gmt":"2020-06-25T06:34:35","slug":"creating-music-by-machine-learning","status":"publish","type":"post","link":"https:\/\/www.aiproblog.com\/index.php\/2020\/06\/25\/creating-music-by-machine-learning\/","title":{"rendered":"Creating Music by Machine Learning"},"content":{"rendered":"<p>Author: Andrew Vo<\/p>\n<div>\n<p><span style=\"font-size: 12pt;\"><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284063683?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284063683?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/span><\/p>\n<p style=\"text-align: center;\"><span style=\"font-size: 12pt;\"><em>&copy; 2020 by Radna Intellectual Ventures, LLC.<\/em><\/span><\/p>\n<p style=\"text-align: center;\">\n<p style=\"text-align: center;\"><span style=\"font-size: 18pt;\"><strong><em>In Sweden, there&#8217;s something called the Allemansr&auml;tten, which literally translates to &#8220;Everyman&#8217;s Right.&#8221;<\/em><\/strong><\/span><\/p>\n<p><span style=\"font-size: 12pt;\"><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284063683?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><\/a><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><strong>Allemansr&auml;tten&nbsp;<\/strong>gives everyone the right to access, trespass, and camp on any land within the immediate vicinity of a residential dwelling.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Imagine a stranger strolling onto your property in the U.S. or U.K., pitching a tent on your lawn, and proclaiming&nbsp;<em>&#8220;Allemansr&auml;tten!&#8221;<\/em>, as they roast marshmallows over an open fire.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">In Sweden, this is legal and they&#8217;re allowed to stay for up to 24 hours.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Allemansr&auml;tten is a very&nbsp;<em>Nordic<\/em><strong>&nbsp;<\/strong>notion of the social contract and it embodies the idea of&nbsp;<strong>Trust&nbsp;<\/strong>that is at the root of Swedish culture.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">The idea of &#8220;public&#8221; vs. &#8220;private&#8221; just doesn&#8217;t exist to some extent in Sweden, and thus it is not surprising that Sweden has been a leader in both&nbsp;<em>Music Piracy<\/em>&nbsp;and&nbsp;<em>Music Innovation<\/em>.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\"><strong>The Age of the Artist<\/strong><\/span><\/h2>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><strong>Music Piracy<\/strong>, with the launch of&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/Napster\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Napster<\/a>, defined much of the 2000s and was catalyzed with the adoption of the&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/MP3\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">MP3<\/a>&nbsp;technology and the launch of&nbsp;<strong>Apple&#8217;s iPod<\/strong>&nbsp;in 2001.<\/span><\/p>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284453874?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284453874?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Getty Images; Swedish DJ Avicii &ndash; the late Tim Bergling &ndash; setting precedence for the upcoming &#8220;Age of the Artist&#8221; of the 2020s.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Even with Napster&#8217;s shut down by the&nbsp;<strong>Recording Industry Association of America<\/strong>&nbsp;(<a href=\"https:\/\/www.riaa.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">RIAA<\/a>) in July 2001, the flood gates were opened and nothing could stop&nbsp;<em>Music Piracy<\/em>, which proliferated in various forms including Sweden&#8217;s own&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/The_Pirate_Bay\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Pirate Bay<\/a>&nbsp;and&nbsp;<a href=\"https:\/\/www.utorrent.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">uTorrent<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">A startling statistic:&nbsp;<a href=\"https:\/\/musicbusinessresearch.wordpress.com\/2015\/03\/26\/the-recorded-music-market-in-the-us-2000-2014\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">CD Album Sales<\/a>&nbsp;topped USD 13.4 Billion in 2000 but by 2014, eroded to just 1.9 Billion, a staggering collapse of&nbsp;<em>86 percent<\/em>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">The 2010s brought with it the exponential growth of&nbsp;<strong>Music Streaming<\/strong>&nbsp;through services like&nbsp;<em>Apple iTunes<\/em>,&nbsp;<em>Amazon Music<\/em>, and&nbsp;<em>Spotify<\/em>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">However, amidst the sea of change in the&nbsp;<strong>Music Distribution<\/strong>&nbsp;industry, the underlying relationship between&nbsp;<em>Artists&nbsp;<\/em>and&nbsp;<em>Labels&nbsp;<\/em>remains relatively unchanged.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">This brings us to the&nbsp;<strong>&#8220;Age of the Artist&#8221;<\/strong>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">The&nbsp;<strong>Age of the Artist<\/strong>&nbsp;represents a&nbsp;<em>structural change<\/em>&nbsp;in the music industry at the&nbsp;<em>&#8220;Top of the Funnel&#8221;<\/em>&nbsp;and will fundamentally shift the relationship between Labels and Artists.<\/span><\/p>\n<\/p>\n<ol>\n<li><span style=\"font-size: 12pt;\"><strong>Alternative Revenue Models<\/strong>: creative deal structures more aligned with Artists than the traditional model of signing artist rights away to a label into perpetuity.<\/span><\/li>\n<li><span style=\"font-size: 12pt;\"><strong>Global Access to Audienc<\/strong>e: according to legendary VC investor&nbsp;<a href=\"https:\/\/www.bondcap.com\/report\/itr19\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Mary Meeker<\/a>, 60% of the world&#8217;s population is connected to the Internet, and with it access to&nbsp;<em>YouTube<\/em>,&nbsp;<em>SoundCloud<\/em>, and a plethora of distribution platforms, allowing any independent artist to share their content and connect with a global audience.<\/span><\/li>\n<li><span style=\"font-size: 12pt;\"><strong>Labels as a Service (&#8220;LaaS&#8221;)<\/strong>:<strong>&nbsp;<\/strong>similar to the idea of&nbsp;<strong>SaaS&nbsp;<\/strong>(&#8220;software as a service&#8221;) &ndash; the subscription-based business model that has disrupted the software industry &ndash; I believe<strong>&nbsp;LaaS<\/strong><em>&nbsp;<\/em>will gain momentum, with Labels becoming more commoditized.<\/span><\/li>\n<\/ol>\n<p><span style=\"font-size: 12pt;\">With the rapid expansion of&nbsp;<em>creation<\/em>,&nbsp;<em>collaboration<\/em>,&nbsp;<em>production<\/em>,&nbsp;<em>marketing,&nbsp;<\/em>and<em>&nbsp;distribution<\/em>&nbsp;tools, the effect is a&nbsp;<strong>shift of bargaining power from Labels to Artists.<\/strong><\/span><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">So the Age of the Artist will be defined by a paradigm shift in companies establishing relationships with artists and songwriters during the creation process, setting up the stage for the new label of the future.<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\"><strong>Spotify<\/strong>&nbsp;&ndash; one of the key disrupters in the&nbsp;<a href=\"https:\/\/www.statista.com\/topics\/1639\/music\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">$22 Billion<\/a>&nbsp;Recorded Music industry &ndash; agrees with this thesis which we will highlight through their&nbsp;<strong>Corporate Development<\/strong>&nbsp;strategy below.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">But before we do, let&#8217;s do a quick primer on&nbsp;<strong>Product Development<\/strong>&nbsp;and specifically the development of products whose core tech is based on&nbsp;<strong>Machine Learning<\/strong>.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\">Product Development in the Age of Machine Learning<\/span><\/h2>\n<\/p>\n<p><span style=\"font-size: 12pt;\">One of my first roles after graduating from Cornell University and U.C. Berkeley with degrees in Computer Science was in Product Management at a financial services firm.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Back in the early 2010s,&nbsp;<strong>&#8220;Data Science&#8221;<\/strong>&nbsp;and&nbsp;<strong>&#8220;AI \/ ML&#8221;<\/strong>&nbsp;were still relatively new words in the financial industry, yet our group leverage&nbsp;<strong>&#8220;Big Data&#8221;<\/strong>&nbsp;and utilized Machine Learning algorithms to estimate movements in certain financial markets with remarkable precision.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">The idea of an&nbsp;<strong>&#8220;AI Product Manager&#8221;<\/strong>&nbsp;simply didn&#8217;t exist back then, yet that was what our group was doing on a daily basis, which allowed us to generate record profits among our competitors in the industry.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Fast forward a decade and the idea of&nbsp;<strong>AI Product Management<\/strong>&nbsp;or&nbsp;<strong>APM<\/strong>&nbsp;is a clearly defined role in Tech, with the largest tech companies like&nbsp;<em>Amazon<\/em>,&nbsp;<em>Facebook<\/em>, and&nbsp;<em>Google&nbsp;<\/em>even having dedicated APM teams.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 10pt;\"><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284560299?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284560299?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/span><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Radna Intellectual Ventures; Traditional Product Development using Wireframes.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>What is the role of an&nbsp;<\/em><strong><em>AI Product Manager<\/em><\/strong><em>?<\/em><\/span><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">The AI Product Manager is principally responsible for using AI, Deep Learning, and Machine Learning to enhance, improve, create, and shape products.<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\"><strong>Artificial Intelligence<\/strong>&nbsp;has far-reaching implications for every conceivable industry according to a&nbsp;<a href=\"https:\/\/www.mckinsey.com\/featured-insights\/artificial-intelligence\/notes-from-the-ai-frontier-modeling-the-impact-of-ai-on-the-world-economy\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">seminal report&nbsp;<\/a>by&nbsp;<em>McKinsey and Company<\/em>, which identified over&nbsp;<strong>400 business use cases of AI<\/strong>&nbsp;for companies today.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Another recent survey of global business leaders and corporate CEOs indicated that over&nbsp;<strong>70% have started AI initiatives<\/strong>. With the proliferation of AI into business (e.g.,&nbsp;<em>Google Search<\/em>,&nbsp;<em>Amazon Product Recommendations<\/em>,&nbsp;<em>Tesla Autopilot<\/em>),&nbsp;<strong>Machine Learning<\/strong>&nbsp;is the common thread among all these novel products.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><a href=\"http:\/\/www.robotics.stanford.edu\/~ang\/contact.html\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Andrew Ng<\/a>&nbsp;&ndash; a Stanford Professor and one of the leading authorities in Machine Learning &ndash; developed the concept of the&nbsp;<a href=\"https:\/\/landing.ai\/content\/uploads\/2020\/05\/LandingAI_Transformation_Playbook_11-19.pdf\" target=\"_blank\" rel=\"nofollow noopener noreferrer\"><strong>&#8220;Virtuous Cycle of AI&#8221;<\/strong><\/a><strong>&nbsp;<\/strong>&ndash; an idea central to the development of Machine Learning products.<\/span><\/p>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284583881?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284583881?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Andrew Ng; Landing AI&#8217;s Playbook featuring the Virtuous Cycle of AI.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Andrew&#8217;s insight that&nbsp;<em>&#8220;building a better AI product requires more users producing more data&#8221;<\/em>&nbsp;has profound implications for the&nbsp;<strong>Product Development Life Cycle.<\/strong><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>Why does this matter?<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Well, products fueled by data and machine learning can be remarkably powerful tools to discover and solve&nbsp;<strong>customer problems<\/strong>, enabling a company to strengthen its&nbsp;<strong>business moat<\/strong>&nbsp;amidst an increasingly competitive environment with lower barriers of entry.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">As the diagram below shows,&nbsp;<strong>Product Development<\/strong>&nbsp;can be broken into four components, which you can read more about in this&nbsp;<em>Harvard Business Review<\/em>&nbsp;<a href=\"https:\/\/hbr.org\/2018\/10\/how-to-build-great-data-products\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">article<\/a>.<\/span><\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284601488?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284601488?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Radna Intellectual Ventures; Product Development in the Age of Machine Learning.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">From the perspective of&nbsp;<strong>AI \/ ML&nbsp;<\/strong>products, here is what these four components really mean.<\/span><\/p>\n<\/p>\n<ol>\n<li><span style=\"font-size: 12pt;\"><strong>User Experience<\/strong>: this is where user problems are discovered and validated. As&nbsp;<em>Marc Andreessen<\/em>&nbsp;of the Venture Capital firm&nbsp;<a href=\"https:\/\/a16z.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">A16Z<\/a>, notes in&nbsp;<a href=\"https:\/\/a16z.com\/2020\/04\/18\/its-time-to-build\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">&#8220;It&#8217;s Time to Build&#8221;<\/a>, innovation is sorely missing in Tech.&nbsp;<strong>UX \/ UI<\/strong>&nbsp;&ndash; beyond providing an attractive and intuitive user interface &ndash; is about discovering what the user really wants and validating the problem, which often leads to&nbsp;<strong>product innovation<\/strong>.<\/span><\/li>\n<li><span style=\"font-size: 12pt;\"><strong>Technical Development<\/strong>: this covers both software or hardware development. Especially relevant for software is the&nbsp;<a href=\"https:\/\/www.agilealliance.org\/agile101\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Agile Software Development<\/a>&nbsp;methodology, which is particularly well-suited for&nbsp;<strong>AI \/ ML product development<\/strong>.<\/span><\/li>\n<li><span style=\"font-size: 12pt;\"><strong>Data Science<\/strong>: although Data Science is a term that is often associated with buzzwords like&nbsp;<strong>&#8220;Cloud Computing&#8221;<\/strong>&nbsp;and<strong>&nbsp;&#8220;Big Data&#8221;<\/strong>, it has been a concrete field used by the best innovators for decades. Leveraging&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/vcs-guide-machine-learning-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">Machine Learning Algorithms<\/a>&nbsp;and Big Data, one can truly achieve the extraordinary &ndash; see&nbsp;<a href=\"https:\/\/www.amazon.com\/Man-Who-Solved-Market-Revolution-ebook\/dp\/B07P1NNTSD\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Jim Simmons<\/a>&nbsp;of&nbsp;<a href=\"https:\/\/www.rentec.com\/Home.action?index=true\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Renaissance Technologies<\/a>&nbsp;&ndash; the most successful investor in history (in fact, twice as good as the legendary Warren Buffet as measured by returns).<\/span><\/li>\n<li><span style=\"font-size: 12pt;\"><strong>Product Management<\/strong>: a Product Manager is akin to the Conductor of an Orchestra. Product Management is an iterative process of gathering user data to build better products, which generates more novel data (see&nbsp;<em>Virtuous Cycle<\/em>&nbsp;above). An exceptional Product Manager, especially in AI \/ ML, is someone with a balance of&nbsp;<strong>technical expertise + business acumen + high customer empathy<\/strong>.<\/span><\/li>\n<\/ol>\n<p><span style=\"font-size: 12pt;\">Given the above, another valuable insight from Andrew Ng is that&nbsp;<strong><em>&#8220;Training Sets are the new Wireframes in product management&#8221;<\/em><\/strong>&nbsp;&ndash; a deep observation which you can read more about in&nbsp;<a href=\"https:\/\/www.gsb.stanford.edu\/insights\/andrew-ng-why-ai-new-electricity\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Why AI is the New Electricity<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">With an understanding of&nbsp;<em>Product Development in the Age of Machine Learning<\/em>, let&#8217;s exercise our knowledge through a&nbsp;<strong>Case Study&nbsp;<\/strong>for one of the most remarkable Machine Learning companies that exist today:&nbsp;<strong><em>Spotify<\/em><\/strong>.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\">Case Study: Spotify<\/span><\/h2>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284635283?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284635283?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\">Source: Getty Images; Spotify&#8217;s Founder and CEO, Daniel Ek.<\/span><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">I&#8217;d argue Spotify&#8217;s history is much more remarkable than Facebook&#8217;s story.<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\">However, largely as a function of its&nbsp;<strong>Swedish&nbsp;<\/strong>origin, Spotify has been less glamorized than&nbsp;<strong>Facebook<\/strong>, as epitomized in the popular American film&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/The_Social_Network\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">&#8220;The Social Network&#8221;<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>What were the humble beginnings of Spotify?<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">It began in the mind of a music-loving young teenager and prodigy,&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/Daniel_Ek\" target=\"_blank\" rel=\"nofollow noopener noreferrer\"><strong>Daniel Ek<\/strong><\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">By the age of nine, Daniel taught himself how to code&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/C%2B%2B\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">C++<\/a>&nbsp;&ndash;&nbsp;<em>one of the more challenging programming languages to master<\/em>&nbsp;&ndash; and wrote 40 pages of code, constructing a cursor to move around the screen.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Daniel continued to hone his coding skills and by age 14 in 1997, Daniel started building websites for other Swedish entrepreneurs, charging a rate of SEK 10,000 per website.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">With this first successful business venture, Daniel caught the&nbsp;<em>entrepreneurial bug,&nbsp;<\/em>planting the seed for the conception and launch of Spotify nine years later in 2006.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">If you don&#8217;t know or use&nbsp;<a href=\"https:\/\/www.spotify.com\/us\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\"><strong>Spotify<\/strong><\/a>, I&#8217;d urge you to&nbsp;<u>stop here<\/u>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><a href=\"https:\/\/support.spotify.com\/us\/using_spotify\/getting_started\/download-the-spotify-player\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Download the app<\/a>&nbsp;on your phone and use it for a week.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">As a&nbsp;<strong><em>Zen<\/em><\/strong>&nbsp;practitioner, I jokingly compare the Spotify music experience to the journey of discovering Zen, as they are both&nbsp;<strong><em>experiential&nbsp;<\/em><\/strong>in nature &ndash; we can talk about both, but you must experience them for yourselves in order to truly understand.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Okay, let&#8217;s move on to the secret sauce behind Spotify, which relies on&nbsp;<strong>Machine Learning<\/strong>.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\">Machine Learning at Spotify<\/span><\/h2>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284671462?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284671462?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Radna Intellectual Ventures; Spotify Paved Road Machine Learning Infrastructure.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Although Spotify has over&nbsp;<strong>50 million songs<\/strong>, it has a staggering&nbsp;<strong>3 Billion Playlists<\/strong>&nbsp;or 60 times more playlists than songs.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>How do we wrap our heads around this?<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">From a Machine Learning perspective, the answer should be clear if you have read the&nbsp;<em>prequel&nbsp;<\/em>to this article:&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/vcs-guide-machine-learning-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">&#8220;The VC&#8217;s Guide to Machine Learning&#8221;<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">If you haven&#8217;t, don&#8217;t panic.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Spotify considers each Playlist as a&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/State-space_representation\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">state space<\/a>&nbsp;consisting of individual states representing a single song. By leveraging&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/Reinforcement_learning\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">reinforcement learning&nbsp;<\/a>techniques and user data, Spotify is able to generate novel &#8220;Journeys&#8221; \/ Playlists through the vast 50-million dimensional&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/Vector_space\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">vector space<\/a>&nbsp;of songs.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>What does that mean in plain English?<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Simply stated, Spotify&#8217;s&nbsp;<strong>content recommendation<\/strong>&nbsp;and&nbsp;<strong>playlist generation<\/strong>&nbsp;rely heavily on state-of-the-art machine learning algorithms to generate remarkably impactful music recommendations for its&nbsp;<a href=\"https:\/\/investors.spotify.com\/financials\/press-release-details\/2020\/Spotify-Technology-SA-Announces-Financial-Results-for-First-Quarter-2020\/default.aspx\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">286 million<\/a>&nbsp;Monthly Active Users (&#8220;MAUs&#8221;).<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Of note, the&nbsp;<strong>development<\/strong>&nbsp;of Machine Learning at Spotify has closely tracked&nbsp;<a href=\"https:\/\/arxiv.org\/list\/cs.LG\/recent\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">academic research<\/a>&nbsp;into&nbsp;<strong>content recommenders<\/strong>&nbsp;and<strong>&nbsp;neural network<\/strong>s and from the early days of&nbsp;<a href=\"https:\/\/www.researchgate.net\/publication\/220616110_A_new_approach_for_combining_content-based_and_collaborative_filters\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Collaborative Filtering<\/a>&nbsp;in 2006 to the latest in&nbsp;<a href=\"https:\/\/arxiv.org\/pdf\/1908.05551.pdf\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Generative Adversarial Networks<\/a>&nbsp;(<strong>GANs<\/strong>), with coherent music generation possible through&nbsp;<strong>LTSM&nbsp;<\/strong>(<a href=\"https:\/\/en.wikipedia.org\/wiki\/Long_short-term_memory\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Long Short-Term Memory<\/a>).<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">You can read more about the&nbsp;<a href=\"https:\/\/engineering.atspotify.com\/2019\/12\/13\/the-winding-road-to-better-machine-learning-infrastructure-through-tensorflow-extended-and-kubeflow\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">technical details&nbsp;<\/a>of how Spotify&#8217;s&nbsp;<strong>&#8220;Paved Road&#8221;<\/strong>&nbsp;machine learning infrastructure is implemented through a collection of open-source software platforms like&nbsp;<a href=\"https:\/\/cloud.google.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Google Cloud<\/a>,&nbsp;<a href=\"https:\/\/www.tensorflow.org\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">TensorFlow<\/a>, and&nbsp;<a href=\"https:\/\/www.kubeflow.org\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Kubeflow<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Let&#8217;s do a quick comparison with&nbsp;<strong>Netflix&nbsp;<\/strong>&ndash; with a market cap of 200 Billion is about 5 times larger than that of Spotify &ndash; whose content recommendation system is also a byproduct of Machine Learning.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">According to this&nbsp;<a href=\"https:\/\/engineering.atspotify.com\/2020\/01\/16\/for-your-ears-only-personalizing-spotify-home-with-machine-learning\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">article<\/a>&nbsp;by Spotify&#8217;s VP of Engineering:&nbsp;<em>&#8220;Netflix has about&nbsp;158 million users, and at Spotify, we have 248 million. On top of that Netflix has around 5,800 movies and shows across all territories, while Spotify has more than 50 million music tracks and 500,000 podcast titles.&#8221;<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">From a&nbsp;<a href=\"https:\/\/en.wikipedia.org\/wiki\/Computational_complexity_theory\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">computational complexity<\/a>&nbsp;perspective, Spotify&#8217;s real-time complexity requirements are an&nbsp;<em>order of magnitude<\/em>&nbsp;larger than that of&nbsp;<strong>NetFlix&nbsp;<\/strong>and quickly growing &ndash; see the observation above about Spotify&#8217;s Playlist generation.<\/span><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">With its Paved Road approach, Spotify has indeed built one of Tech&#8217;s most robust and efficient Machine Learning infrastructure.<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\">Let&#8217;s now shift gears from&nbsp;<em>Product to Business<\/em>.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\">Corporate Development at Spotify<\/span><\/h2>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><strong>Corporate Development<\/strong>&nbsp;is an ambiguous industry term, but generally for larger companies, it relates to activities like acquiring other companies (<em>M&amp;A<\/em>), securing&nbsp;<em>corporate financing<\/em>,&nbsp;<em>divesting<\/em><strong><em>&nbsp;<\/em><\/strong>of assets or divisions, and management of the&nbsp;<em>intellectual property (IP)<\/em>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Among the&nbsp;<strong>Tech Decacorns<\/strong>&nbsp;(companies with a valuation greater than $10 Billion), Spotify has perhaps one of the most active Corporate Development teams in the industry, with a large part of that activity focused on acquiring startups with novel music tech.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">In a similar vein to what we did in&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/vcs-guide-machine-learning-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">&#8220;The VC&#8217;s Guide to Machine Learning&#8221;<\/a>, we have mapped Spotify&#8217;s startup acquisitions with&nbsp;<em>investable themes&nbsp;<\/em>in the&nbsp;<strong>AudioTech&nbsp;<\/strong>industry.<\/span><\/p>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284724082?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284724082?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<p><span style=\"font-size: 12pt;\"><em>Source: Radna Intellectual Ventures; Investable Themes in AudioTech: Connecting Spotify&#8217;s Strategic Vision with their Acquisition Strategy.<\/em><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Although music continues to be their core offering, Spotify&#8217;s strategic vision is to expand into&nbsp;<strong>Audio<\/strong>&nbsp;in general like&nbsp;<strong>Podcasting<\/strong>&nbsp;&ndash;<strong>&nbsp;<\/strong>a thesis reinforced by their numerous acquisitions of Podcasting startups like&nbsp;<a href=\"https:\/\/anchor.fm\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Anchor<\/a>,&nbsp;<a href=\"https:\/\/www.parcast.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Parcast<\/a>, and&nbsp;<a href=\"https:\/\/gimletmedia.com\/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Gimlet<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">From a thematic perspective, I expect Spotify&#8217;s leadership &ndash; principally driven by Daniel Ek and his Lieutenant&nbsp;<a href=\"https:\/\/twitter.com\/gustavs\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Gustav S&ouml;derstr&ouml;m<\/a>&nbsp;&ndash; to shift more investment capital to&nbsp;<strong>Content Distribution \/ Licensing \/ Marketing<\/strong>&nbsp;and &#8220;Exploration&#8221;.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">I define&nbsp;<strong>Exploration&nbsp;<\/strong>broadly as tools and services that enable music (and audio in general) listeners to (1) discover new content and (2) experience audio content in a&nbsp;<strong>novel manner<\/strong>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Novel in what way?<\/span><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">With the coming of age of Virtual Reality and Augmented Reality (which is a more challenging problem from a technical perspective), I can see virtual concerts and live &#8220;jam sessions&#8221; being an active area of investment for Spotify this decade.<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\">Returning now to&nbsp;<strong>Allemansr&auml;tten&nbsp;<\/strong>where we started.<\/span><\/p>\n<\/p>\n<h2><span style=\"font-size: 18pt;\">Final Thoughts<\/span><\/h2>\n<\/p>\n<p><span style=\"font-size: 12pt;\">As you may recall&nbsp;<strong>Allemansr&auml;tten&nbsp;<\/strong>or &#8220;Everyman&#8217;s Right&#8221; is a core tenet of Swedish culture and I believe it is also a core tenet of&nbsp;<strong>Decentralization<\/strong>&nbsp;during this&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/aidos-global-trends-roaring-2020s-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">Roaring 2020s<\/a>.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Decentralization &ndash; with&nbsp;<strong>Blockchain&nbsp;<\/strong>as the enabling technology &ndash; will play a key role during this&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/ventures-deep-technologies-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">Fourth Industrial Revolution<\/a>, and unless you get on board the &#8220;Allemansr&auml;tten train&#8221;, you will very likely be missing out on the next Trillion dollar opportunity in Tech.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">So if you enjoy the scenic train view, let&#8217;s continue&nbsp;<strong>Creating Music by Machine Learning<\/strong>.<\/span><\/p>\n<\/p>\n<p><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284757896?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/6284757896?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/p>\n<\/p>\n<blockquote><p><span style=\"font-size: 18pt;\">A good composer does not imitate; he steals.<\/span><\/p><\/blockquote>\n<blockquote><p><span style=\"font-size: 18pt;\">&mdash;Igor Stravinsky<\/span><\/p><\/blockquote>\n<p><span style=\"font-size: 12pt;\"><strong>By Andrew Vo, CFA<\/strong><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Investor at Radna Intellectual Ventures<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><em>Related Link:<\/em>&nbsp;<a href=\"https:\/\/www.linkedin.com\/pulse\/vcs-guide-machine-learning-andrew-vo-cfa\/\" target=\"_blank\" rel=\"noopener noreferrer\">The VC&#8217;s Guide to Machine Learning<\/a><\/span><\/p>\n<\/p>\n<h3><span style=\"font-size: 14pt;\">About Radna Intellectual Ventures<\/span><\/h3>\n<\/p>\n<p><span style=\"font-size: 12pt;\"><strong>Radna Intellectual Ventures<\/strong>&nbsp;is a venture startup studio with a mission to enable the next generation of Deep Technologies or Deep Tech companies, with a focus on Machine Learning, Natural Language Processing, and Blockchain technologies.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 10pt;\"><a href=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/5494972473?profile=original\" target=\"_blank\" rel=\"noopener noreferrer\"><img decoding=\"async\" src=\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/5494972473?profile=RESIZE_710x\" width=\"750\" class=\"align-center\"><\/a><\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Our General Partner,&nbsp;<strong>Andrew Vo<\/strong>, spent a decade in finance, working at some of the largest investment managers like J.P. Morgan, before pursuing the entrepreneurial path. Andrew serves as an Advisor for startup companies focusing on the FinTech and DeepTech sectors. Andrew is a CFA Charterholder and holds a Masters of Science in Computer Engineering from Cornell University, and a Bachelors of Science in Electrical Engineering and Computer Science from the University of California, Berkeley.<\/span><\/p>\n<\/p>\n<p><span style=\"font-size: 12pt;\">Learn more about&nbsp;<strong>Radna Intellectual Ventures<\/strong>&nbsp;<a href=\"https:\/\/www.linkedin.com\/company\/51702405\/\" target=\"_blank\" rel=\"noopener noreferrer\">here<\/a>.<\/span><\/p>\n<\/p>\n<div id=\"ember8792\" class=\"ember-view\">\n<div dir=\"ltr\" class=\"reader-article-content\">\n<p><span style=\"font-size: 12pt;\">&mdash; Published on&nbsp;June 24, 2020<\/span><\/p>\n<\/div>\n<\/div>\n<div class=\"reader-flag-content__wrapper mb4 clear-both\"><\/div>\n<\/p>\n<\/div>\n<p><a href=\"https:\/\/www.datasciencecentral.com\/xn\/detail\/6448529:BlogPost:959946\">Go to Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Author: Andrew Vo &copy; 2020 by Radna Intellectual Ventures, LLC. In Sweden, there&#8217;s something called the Allemansr&auml;tten, which literally translates to &#8220;Everyman&#8217;s Right.&#8221; Allemansr&auml;tten&nbsp;gives everyone [&hellip;] <span class=\"read-more-link\"><a class=\"read-more\" href=\"https:\/\/www.aiproblog.com\/index.php\/2020\/06\/25\/creating-music-by-machine-learning\/\">Read More<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":457,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_bbp_topic_count":0,"_bbp_reply_count":0,"_bbp_total_topic_count":0,"_bbp_total_reply_count":0,"_bbp_voice_count":0,"_bbp_anonymous_reply_count":0,"_bbp_topic_count_hidden":0,"_bbp_reply_count_hidden":0,"_bbp_forum_subforum_count":0,"footnotes":""},"categories":[26],"tags":[],"_links":{"self":[{"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/posts\/3598"}],"collection":[{"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/comments?post=3598"}],"version-history":[{"count":0,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/posts\/3598\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/media\/467"}],"wp:attachment":[{"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/media?parent=3598"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/categories?post=3598"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.aiproblog.com\/index.php\/wp-json\/wp\/v2\/tags?post=3598"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}